Photoelectric Conversion Filters With Local AI Models for Faster Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing photoelectric conversion apparatuses face processing speed limitations and increased load due to a single processing unit performing signal processing based on a learned model, as described in Japanese Patent Laid-Open No. 2020-25263.
Innovation Solution
The apparatus is designed with multiple AI processing units on a second substrate, each processing outputs from photoelectric conversion units with different color filters, allowing parallel processing and optimized learned models for each filter type, and utilizing Chip-on-Chip, Chip-on-Wafer, or Wafer-on-Wafer bonding methods for electrical connection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single processing unit performs signal processing based on a learned model, then the apparatus structure is simplified, but processing speed decreases and the processing load increases
Solution Approach 1:
The patent divides the single processing unit into multiple AI processing units (first AI processing unit, second AI processing unit, etc.), each capable of independently performing signal processing based on learned models. This segmentation enables parallel processing of signals from different photoelectric conversion units, thereby increasing overall processing speed and reducing the processing load on each individual unit while maintaining relatively simple apparatus structure.
2Device complexity
If a single processing unit performs signal processing, then device complexity is reduced, but heat generation increases and power consumption rises
Solution Approach 1:
By segmenting the processing function across multiple AI processing units, the patent distributes heat generation across multiple components rather than concentrating it in a single unit. This dispersion of thermal load reduces the temperature increase in any one location and allows for better heat management, while still achieving the desired signal processing functionality with relatively simple apparatus structure.
3Productivity
If multiple AI processing units with different learned models are used, then processing speed and image quality are enhanced, but device complexity increases
Solution Approach 1:
The patent applies local quality by assigning different learned models to different AI processing units based on their specific functions. Each AI processing unit is optimized with a learned model tailored to its particular processing requirements (e.g., different color filters, different processing tasks), thereby achieving high processing speed and image quality for each specific function while maintaining overall apparatus simplicity through this targeted, localized optimization approach.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This configuration enhances processing speed, reduces heat generation and power consumption, and optimizes image quality by performing parallel and dispersed processing, minimizing temperature increases and maintaining high accuracy.
Implementation Method 1
a photoelectric conversion region 101 in which a plurality of photoelectric conversion units are arranged
Data Source
AI summary
A photoelectric conversion apparatus has a first filter arranged so as to correspond to a first photoelectric conversion unit and a second filter arranged so as to correspond to a second photoelectric conversion unit. The photoelectric conversion apparatus has a first processing unit configured to process an output signal from the first photoelectric conversion unit and having a first learned model, and a second processing unit configured to process an output signal from the second photoelectric conversion unit and having a second learned model different from the first learned model.